⏰ OpenRouter Limited-Time Free Models
Two limited-time free models expire tomorrow, June 29, 2026:
- Anthropic: Claude Opus 4.6 (Fast) — 1M context, $30/M tokens prompt / $150/M tokens completion. One of the strongest reasoning models available for free testing.
- StepFun: Step 3.5 Flash — 262K context, $0.09/M tokens prompt / $0.30/M tokens completion. A lightweight, cost-efficient flash model ideal for high-volume workloads.
Both models reach their final free day tomorrow — evaluate them before the window closes.
🤗 Trending on Hugging Face
This batch highlights a strong Arabic/regional-language theme and specialized models:
- BinSaqban/Qwen2.5-7B-Instruct-Arabic — The most downloaded entry at 835 downloads, fine-tuned for Arabic instruction following.
- BinSaqban/Averroes-Q — A Qwen2-based Arabic model from the Hayula/Agent-OS ecosystem, part of a broader push for regional-language AI.
- BinSaqban/Hayula-Rushd-Q-Instruct-v4 — Latest iteration of the Rushd-Q instruct series with human-curated Arabic training data.
- BinSaqban/Averroes-Q-Instruct-human55k — 55K human-annotated Arabic instructions for fine-tuning Qwen2 models.
- BinSaqban/Hayula-Rushd-Q-v1 — Companion base model to the instruct variant.
- nerkyor/Qwen3.6-27B-DSV4Pro-Thinking-Distill-NVFP4 — A reasoning-focused distillation of Qwen3 for the DeepSeek V4 architecture with NVFP4 quantization. ⭐ 1
- weareapexcreators/Granite-4.1-3B-LocalAI — IBM’s Granite 4.1 3B model packaged for LocalAI with GGUF format and imatrix support.
- Dama12/dioula-tts-vits — VITS-based text-to-speech model for the Dioula language (West Africa), built on Coqui TTS.
- SpeechAntiSpoofingBenchmarks/RawBMamba — ONNX-deployable anti-spoofing model for audio authenticity detection.
- chilkersion/KurutAI — A new community model gathering traction. ⭐ 3
⭐ GitHub Trending: AI Edition
- winsznx/theeleven — Eleven autonomous AI agents running live football prop markets on X Layer via a custom Uniswap v4 hook. A fascinating intersection of AI agents, DeFi, and sports betting. ⭐ 645 — TypeScript
- benchflow-ai/awesome-evals — A curated library of the best resources for evaluating AI agents — papers, blogs, and frameworks. Essential bookmark for anyone building production agents. ⭐ 537
- Pluviobyte/video-production-skills — Reusable AI video production skill library for creation, recreation, motion design, and QA workflows. ⭐ 339 — Python
- trotsky1997/OpenFugu — Open reimplementation of Sakana Fugu — the “one model to command them all” LLM orchestrator. Reflects surging interest in meta-agent architectures. ⭐ 267 — Python
- eli-labz/Godcoder — A local-first, open-source coding agent for your desktop. Bring your own LLM key; your code stays on your machine. ⭐ 240 — Rust
💡 Key Trends
-
AI for the Workforce: $500M RAISE US Initiative. OpenAI, Anthropic, Microsoft, and Amazon jointly committed $500M to the RAISE US workforce retraining fund, targeting retraining of American workers for an AI-driven economy — with a goal of reaching $1B. This marks the first major cross-industry coalition on AI workforce transition, signaling that frontier labs recognize the labor-market disruption their products will create.
-
Hardware Talent War Heats Up. Apple Vision Pro executive Paul Meade is reportedly leaving for OpenAI’s hardware team, signaling OpenAI’s deepening commitment to custom hardware beyond the “Jalapeño” inference chip with Broadcom. Meanwhile, ON Semiconductor struck a $7B deal for Synaptics in a physical AI push, and ElevenLabs adopted Google’s SynthID watermarking for AI-generated audio.
-
Anthropic’s Rollercoaster Week. After Mythos 5 was pulled offline amid export-control tensions with the Trump administration, the model has now been restored following negotiations. Anthropic’s valuation is approaching $1T as the company prepares for a potential IPO, while simultaneously dropping workplace AI agents directly inside Slack for enterprise customers.
-
Amazon Prime Day 2026: AI Delivers 40% Better Conversion. Adobe data shows AI-directed shoppers converted at 40% higher rates than non-AI channels during Prime Day 2026, with total online sales hitting $26.4B (up 9.3% YoY). This is a concrete ROI milestone for AI in e-commerce — measurable, attributable, and large-scale.
-
LLM Orchestration and Open-Source Agents Surge. OpenFugu’s open reimplementation of Sakana Fugu’s “one model to command them all” architecture, alongside eli-labz/Godcoder’s local-first coding agent, reflects a surge in open-source meta-agent and autonomous coding tooling. Google also launched its Interactions API as the primary interface for Gemini models and agents.
Today’s Snapshot
- Trending HF Models: 10
- New OpenRouter Models: 0
- Trending GitHub Repos: 5
- Limited-Time Free Models Expiring: 2 (Claude Opus 4.6 Fast, Step 3.5 Flash — ends June 29)
- Total Items: 17
- News Highlights: Amazon Prime Day AI conversion, RAISE US $500M fund, Anthropic Mythos 5 restored, Apple exec → OpenAI hardware, ON Semi $7B Synaptics deal